This paper provides a structural empirical analysis of Dutch auctions of houseplants at the flower auction in Aalsmeer, the Netherlands. The data set is unique for Dutch auctions in the sense that it includes observations of all losing bids in an interval adjacent to the winning bid. The size of this interval is determined by the speed of reaction of the auction participants, and as such these data are collectible due to neurological constraints on information processing. The data on losing bids are shown to be informative on the structural model determinants. The models are estimated using the Gibbs sampler with data augmentation. We take account of data limitations concerning the number of bidders. The estimation results are used to investigate whether actual reserve prices are optimal, and to determine the effects of reserve price changes.